Mobile robot localization using incomplete maps for change detection in a dynamic environment

Summary form only given. We have developed a system in which a mobile robot can modify maps of an indoor environment automatically through change detection at planned locations. The robot is a car-like one equipped with sonar sensors. The indoor space is assumed to be surrounded by walls and has some movable objects of planar surfaces perpendicular to the floor. In the map updating process, the robot moves to a planned observation location and collect range data on the local region around itself by using the sonar range sensors. The robot's position is estimated by a localization method, and visible changes are detected by comparison of structural range descriptions with the original and partially updated maps. As the result, the map is further modified and a new observation location is selected. In this paper, we present a method by which a mobile robot estimates its position at each location it reaches by using information from dead reckoning, real sonar range data and the partially outmoded map. Simulation experiments show the method is efficient enough for localization in a changeable complex environment.